Projects

Project · Completed · Sep 2026 — Oct 2026

Personal Portfolio & Research Archive

A static engineering portfolio and research archive designed around distinct reader needs, typed content, reproducible delivery and progressively enhanced interaction.

Role
Product & Software Engineering
Period
Sep 2026 — Oct 2026
Status
Completed
Stack
Astro · TypeScript · MDX / Content Collections · GitHub Actions · GitHub Pages
Outcome
A deployed engineering portfolio and research archive with a custom domain, typed content collections, project and publication case studies, deterministic citation tooling, bounded client-side interaction, automated validation and GitHub Pages deployment.
Light desktop and dark mobile views of sergeygalichenko.dev showing the Home interface and interactive system topology

Context and requirements

The project started with a broader question than how to build a personal website: what information should different readers be able to find, how quickly should they find it, and what evidence should remain available when they need more depth?

I treated hiring managers, technical leads, researchers, faculty and engineering colleagues as distinct readers. That led to an identity-first Home page, selected work as the primary evidence, separate Projects, Research and Notes archives, and detailed pages for engineering and scholarly material.

This requirements-first approach shaped the information architecture before implementation rather than allowing the framework or visual design to determine the structure.

Content architecture

Projects, publications and notes are stored as structured content rather than embedded directly into page components. Astro Content Collections and TypeScript schemas validate their metadata and allow new records to reuse the same archive, detail-page, linking and metadata infrastructure.

Project pages focus on the problem, role, engineering decisions, implementation and evidence. Publication records carry richer scholarly metadata including authorship, status, venue, identifiers, abstracts, keywords, related projects and available resources.

The structure also anticipates localization without requiring a second implementation of the site. English is the current public language.

Projects and Research archive interfaces with the citation tool for structured publication content
Projects, research records and citation tooling are generated from structured content.

Static system architecture

Astro and TypeScript generate the site as static output. GitHub Pages serves that output on the custom sergeygalichenko.dev domain.

The architecture deliberately avoids a runtime application backend, database, CMS and user-account system because none is required for the current product. Keeping the runtime surface small reduces operational complexity and attack surface while making the public site inexpensive to host and straightforward to reproduce.

Client-side code is reserved for progressive enhancements such as navigation behavior, theme handling, citation interaction and the interactive system topology on the Home page.

A small resilient system model

The Home-page topology began as a visual counterweight to the identity block and developed into a bounded client-side system model.

Normal traffic is routed across processing paths; repeated interaction introduces bounded concurrency, admission control and backpressure. Sustained deliberate load can isolate one processing path, trigger a single failover, and later return the path only after a health probe. Observability activity is aggregated separately and can expose the trace of the latest completed transaction.

The model is intentionally local and finite: it performs no network requests, keeps queues and transient SVG work bounded, and cannot produce a total simulated outage. It is an interaction detail rather than a claim that the portfolio itself requires distributed infrastructure.

Research and publication tooling

Research content has requirements that ordinary portfolio cards do not. Publication records distinguish bibliographic metadata, local files, publisher resources and scholarly records instead of treating them as interchangeable links.

The citation tool generates GOST R 7.0.5–2008, IEEE, APA 7 and BibTeX representations deterministically from the stored publication metadata. No external scholarly API is required at runtime.

The same structured records also drive archive pages, detail routes, relationships between projects and publications, metadata and sitemap output.

Validation and delivery

The repository keeps application code, content and deployment configuration together. GitHub Actions validates and builds the static output before deployment to GitHub Pages.

The development workflow uses linting, strict TypeScript checks, production builds and focused output tests. Separate audit tooling can generate reproducible review bundles containing rendered output, metadata and selected screenshots without changing the production site.

Codex was used extensively as an implementation assistant. I defined the requirements, architecture, interaction behavior and acceptance criteria, reviewed the resulting changes and decided what entered the production branch.

Outcome

The first standalone version was completed in October 2026.

The result is a deployed engineering portfolio and research archive with a custom domain, typed content model, project and publication case studies, deterministic citation tooling, responsive light and dark themes, a bounded interactive Hero system, automated validation and GitHub Pages deployment.

The implementation phase is complete, while the archive itself is expected to keep growing as new engineering work, publications and notes are added.